Parkinson’s disease is the second most common neurodegenerative disorder among the elderly worldwide and often leads to motor function impairments. In clinical diagnosis, Toe-Tapping is an important assessment indicator that reflects the severity of motor symptoms such as tremor and rigidity. However, traditional clinical assessments are often subjective and inefficient. Currently, research using computer vision to analyze Toe-Tapping in Parkinson’s patients remains limited, and there is an urgent need to establish benchmarks. To address this, we built a large-scale, multi-view video dataset of Toe-Tapping in Parkinson’s patients using three cameras. The dataset includes 1026 video clips from 176 participants. We evaluated five types of mainstream video-based action recognition models to explore the feasibility of using computer vision algorithms to analyze Toe-Tapping movements as a reference for clinical applications. Furthermore, we summarize the current challenges and opportunities in this field, aiming to provide both data resources and methodological references for future research.